mathewthe2 / Game2Text

Complete toolbox for gamifying language learning
https://www.Game2Text.com
Apache License 2.0
186 stars 24 forks source link

OCR doesn't work properly when text is not white #27

Closed blacktide082 closed 3 years ago

blacktide082 commented 3 years ago

For some reason when the text is a different color the OCR generates gibberish. This is in AI: The Somnium Files.

game2text1

game2text2

White text appears to work fine though:

game2text3

blacktide082 commented 3 years ago

This is using the default settings, with Tesseract Default. Interestingly all of the OCR options give the same gibberish except for OCR Space EU which returns the correct text.

mathewthe2 commented 3 years ago

Have you checked the manual on improving OCR?

https://game2text.com/user-guide/improve-accuracy/apply-image-filters/

Also, since you are on Windows, you should use the text hooker functionality for this game.

https://game2text.com/user-guide/visual-novels/

blacktide082 commented 3 years ago

Yes I use text extraction for visual novels, but I am testing out the OCR since I suggested using this to someone else for playing games that aren't visual novels. Thanks for the info. I'll try using the image filters.